Conventionally, compression of information representing the attribute of each pixel which forms an image, i.e., so-called attribute information uses lossless coding so as not to change the attribute information. However, the compression ratio of lossless coding changes depending on the contents of data to be coded (in this case, attribute information).
When attribute information including a lot of information (attribute information formed from multiple bits) is coded, if a bit value at the same position is referred to, the appearance frequency of the same bit value is low. Hence, it has been difficult to obtain a desired coding efficiency by only simple predictive coding.